Utilizing genetic code expansion to modify N-TIMP2 specificity towards MMP-2, MMP-9, and MMP-14

Matrix metalloproteinases (MMPs) regulate the degradation of extracellular matrix (ECM) components in biological processes. MMP activity is controlled by natural tissue inhibitors of metalloproteinases (TIMPs) that non-selectively inhibit the function of multiple MMPs via interaction with the MMPs’ Zn2+-containing catalytic pocket. Recent studies suggest that TIMPs engineered to confer MMP specificity could be exploited for therapeutic purposes, but obtaining specific TIMP-2 inhibitors has proved to be challenging. Here, in an effort to improve MMP specificity, we incorporated the metal-binding non-canonical amino acids (NCAAs), 3,4-dihydroxyphenylalanine (L-DOPA) and (8-hydroxyquinolin-3-yl)alanine (HqAla), into the MMP-inhibitory N-terminal domain of TIMP2 (N-TIMP2) at selected positions that interact with the catalytic Zn2+ ion (S2, S69, A70, L100) or with a structural Ca2+ ion (Y36). Evaluation of the inhibitory potency of the NCAA-containing variants towards MMP-2, MMP-9 and MMP-14 in vitro revealed that most showed a significant loss of inhibitory activity towards MMP-14, but not towards MMP-2 and MMP-9, resulting in increased specificity towards the latter proteases. Substitutions at S69 conferred the best improvement in selectivity for both L-DOPA and HqAla variants. Molecular modeling revealed how MMP-2 and MMP-9 are better able to accommodate the bulky NCAA substituents at the intermolecular interface with N-TIMP2. The models also showed that, rather than coordinating to Zn2+, the NCAA side chains formed stabilizing polar interactions at the intermolecular interface with MMP-2 and MMP-9. The findings illustrate how incorporation of NCAAs can be used to probe and exploit differential tolerance for substitution within closely related protein-protein complexes to achieve improved specificity.


Introduction
Human matrix metalloproteinases (MMPs) comprise a family of 28 known zinc-dependent catalytic enzymes that play major roles in the degradation of extracellular matrix (ECM) components. The role of MMPs in ECM remodeling renders them signi cant players in biological processes as diverse as angiogenesis, tissue hemostasis, wound healing and embryogenesis 1 . MMPs are thus attractive as therapeutic targets, particularly since imbalances in MMP activity or expression can promote pathological conditions, such as arthritis, cardiovascular diseases, and cancer progression, invasion and metastasis [1][2][3][4] . For example, among the MMP family, MMP-2, MMP-9 and MMP-14 are expressed in 70-100% of invasive breast tumors. The differential expression of these MMPs as markers for different diseases, including different cancers, and the ability of this expression to change over time highlight the importance of developing tailored therapeutic strategies for their selective inhibition 5 . Although belonging to different subgroups of MMPs, the gelatinases, MMP-2 and MMP-9, and the membrane-type MMP, MMP-14, exhibit high similarity in their sequences and structures. This similarity is paradoxically both an advantage and a disadvantage; the former because their X-ray structures have been solved and are available for bioinformatic analysis of their interactions with each other and with their inhibitors and substrates 6 , and the latter because the design of speci c inhibitors becomes a very challenging task.
One of the structural characteristics that is common to all MMP family members is the catalytic domain, which has a conserved zinc-binding motif, HEXGHXXGXXH, a catalytic Zn 2+ ion, a structural Zn 2+ ion, and two (or three) Ca 2+ structural ions, all of which contribute to its stabilization. The catalytic Zn 2+ ion is coordinated to three His residues of the conserved motif and to one water molecule. When the catalytic domain binds to a substrate, the coordinated water molecule becomes polarized between the conserved catalytic glutamate base in the zinc-binding motif and the catalytic Zn 2+ Lewis acid to facilitate a nucleophilic attack on a peptide bond, resulting in substrate hydrolysis 7 . Substrate speci city is determined by the size and shape of the six contact sites surrounding catalytic Zn 2+ ion 1,8−10 . Of these, the S1 speci city pocket differs most in size, shape and amino acid content among the different MMPs, and substrates (or inhibitors) with complementary properties at this site may exhibit higher a nity and selectivity for particular MMPs over others.
The interaction of MMPs with the four tissue inhibitors of metalloproteinases (TIMP1-TIMP4) is an important mechanism by which MMP activity is regulated in vivo 11,12 , with the major TIMP-MMP interaction taking place through the binding of the N-terminus of the TIMP (Cys1-Pro5) to the S1, S1 , S2 , S3 , and S4 pockets of the MMP 13 . Speci cally, the Cys1 residue (which is bound to Cys72 via a disul de bond) interacts with the catalytic Zn 2+ ion of the MMP via its N-terminal α-amino and carbonyl groups, displacing the water molecule and thus serving as a fourth zinc ligand. Cys1 also interacts with the catalytic glutamic acid residue of the zinc-binding motif via a hydrogen bond. The second residue of the TIMP, being either threonine or serine 14 , can also form a hydrogen bond with the catalytic glutamic acid, as may be seen in some, but not necessarily all, crystal structures 15 . Several studies have thus set out to manipulate this MMP-TIMP interaction as a means of improving its speci city, for example, by replacing Ser2 with Glu or Asp20 orSer68 with Arg21 16,17 . In particular, considerable effort has been devoted to developing high-a nity inhibitors with good speci city for particular MMPs for therapeutic applications. However, the potential of most synthetic MMP inhibitors 9,18−20 that were designed to chelate the catalytic Zn 2+ ion has not been realized: these synthetic compounds exhibit good inhibition activity, but their selectivity is limited and they are not suitable for clinical use due to poor solubility 21 , poor pharmacokinetics, low bioavailability and severe adverse effects 22,23 . In contrast, TIMP2 and its isolated N-terminal domain, N-TIMP2, have shown promising inhibition potency towards the MMP family in that they exhibit high a nity toward various MMPs (10 − 12 -10 − 9 M), and -being native human proteins -they are likely to be non-toxic and non-immunogenic. However, although N-TIMP2 shows high a nity to MMPs, it lacks speci city for particular MMPs. We therefore sought to improve the speci city by using genetic code expansion, in which non-canonical amino acids (NCAAs) are site-speci cally incorporated into target proteins. The rationale for this approach is based on previous studies in which NCAAs were used to improve the selectivity of different peptides and proteins. For example, 4-tert-butyl and 4-aminomethyl derivatives of phenylalanine were shown to have 20-to 30-fold higher a nity than phenylalanine for their synthetic receptor (Q7), respectively 24 . Previous studies have also shown that the incorporation of NCAAs into proteins and small peptides can be used to improve and to tune their metal binding a nities 25 , the importance of which derives from the crucial roles of metal ions in many biological processes, such as apoptosis, oxidative stress, and immune defense.
Here, we sought to incorporate NCAAs into N-TIMP2 so as to differentially modulate inhibition of different MMPs in a manner that would enhance speci city. We leveraged the NCAAs to probe the potential of bulky polar residues with metal-binding capability either to differentially enhance MMP binding or to be differentially tolerated by different MMPs. The two metal-binding NCAAs that we judged to be suitable for this study were 3,4-dihydroxyphenylalanine, also known as L-DOPA or hydroxytyrosine, which has a catechol side chain, and (8-hydroxyquinolin-3-yl)alanine (HqAla), which has a derivative of the 8hydroxyquinoline chelating moiety in its side chain. Our choice of these two NCAAs was based on previous studies showing that catechol-based compounds 26-28 and 8-hydroxyquinoline derivatives 29,30 exhibited promising potential as inhibitors of MMP-2, MMP-9 and MMP-14, with IC 50 values in the low and even the sub-micromolar range, and that they displayed anti-MMP activity in proliferation, migration and zymography assays. Furthermore, previous studies showed that, when incorporated into a small peptide, L-DOPA exhibited zinc binding consistent with a 1:1 peptide:zinc complex 25 , and when incorporated into alcohol dehydrogenase II it facilitated an increase in Zn 2+ binding, compared to the wild-type protein 31 . It was also shown that HqAla binds divalent ions, such as Zn 2 + 32,33 , Cu 2+ 33 and Ca 2 + 34 , and that incorporation of HqAla into different proteins increased the metal-binding capabilities of those proteins.
In this study, we thus used genetic code expansion to incorporate the bulky, polar, metal-binding NCAAs L-DOPA and HqAla into various positions in N-TIMP2 that are located near the catalytic Zn 2+ ion or to one of the Ca 2+ structural ions of a bound MMP. We reasoned that incorporation of a metal-binding NCAA into N-TIMP2 could increase the ability of the mutant N-TIMP2 to chelate the Zn 2+ or Ca 2+ ions in the catalytic domain and thereby disrupt the catalytic activity of the MMP. Furthermore, the use of a metal-binding NCAA with bulky polar residues might be expected to endow selective a nity of the NCAA-N-TIMP2 towards some MMPs in preference to others, due to differences in the catalytic domain subsites that are crucial for substrate/inhibitor binding.

Results
Choosing positions in N-TIMP2 for site-speci c incorporation of NCAAs. In an attempt to enhance the speci city of N-TIMP2, we chose a strategy that rests on site-selected incorporation of a single NCAA, either L-DOPA or HqAla, at the MMP-binding interface (Fig. 1A). The thinking underlying this strategy was to exploit the steric factor that comes into play when natural amino acids are replaced with bulky NCAAs. In applying the chosen strategy, we targeted several positions in N-TIMP2 that are not only located close to the MMP's Zn 2+ and Ca 2+ ions when the two proteins interact (to potentially take advantage of the metal-binding capability of the NCAA) but also interact with MMP subsites that are not highly conserved. The mutated N-TIMP2 subsites include residues S2, Y36, S69, A70 and L100 (Fig. 1B). To incorporate the NCAAs into N-TIMP2, suppression of the amber codon was performed by introducing (using PCR) a TAG codon at each of the selected positions in N-TIMP2 (one in each clone).
Incorporation of NCAAs into N-TIMP2. For amber suppression and incorporation of L-DOPA or HqAla into different positions in N-TIMP2, we co-transformed Escherichia coli strain WK6 with two plasmids, as described in the Methods section. Wild-type N-TIMP2 and the N-TIMP2-DOPA and N-TIMP2-HqAla variants were produced in the bacteria and puri ed using a nity chromatography (Fig. 2). Mass spectrometry con rmed the successful incorporation of L-DOPA and HqAla at each one of the selected N-TIMP2 positions. Speci cally, MALDI-TOF analysis showed the expected mass difference due to each substitution, compared to N-TIMP2 ( Fig. 3A Table 1). None of the variants containing NCAAs showed improved inhibition toward any of the MMPs tested. However, as intended, the substitutions diminished the inhibitory activity toward the different MMPs (although to widely varying extents), resulting in an enhancement of speci city. Whereas N-TIMP2 bound MMP-14 CAT with a K i of 0.71 nM (Table 1), a nding consistent with previous studies 35,36 , nearly all the N-TIMP2-DOPA and N-TIMP2-HqAla mutants lost their inhibitory activity towards MMP-14 CAT by more than one order of magnitude compared to N-TIMP2. In contrast, most of the N-TIMP2-DOPA and N-TIMP2-HqAla mutants retained their inhibition potency towards MMP-2 ACT and MMP-9 CAT . Notably, N-TIMP2-Y36HqAla, N-TIMP2-S69HqAla and N-TIMP2-S69DOPA exhibited the best inhibition of MMP-2 ACT and MMP-9 CAT , with only a one-to twofold diminishment of potency compared to N-TIMP2. The MMP inhibition assays suggested that the selected positions within N-TIMP2 exhibit different degrees of tolerance for mutagenesis -induced by either L-DOPA or HqAla -that differentially impact their potency toward the different MMPs.  Figure 5 to Morrison's tight binding equation. b K i (fold) is calculated as the ratio between the K i of N-TIMP2 variant and the K i of N-TIMP2.     of the MMP (Fig. 6), suggesting that Ser69 may be a signi cant determinant of a nity toward MMP-14.
With the exception of this speci c interaction, the complexes were overall very similar. We also observed that a nearby MMP residue at the equivalent position 196/193/204 (numbering from initiator methionine for all enzymes) was not conserved, being alanine in MMP-2, proline in MMP-9, and phenylalanine in MMP-14.
The modeling with the HqAla variants revealed the importance of MMP sequence differences at positions 196/193/204 (Fig. 7A). In both MMP-2-and MMP-9-bound complexes, HqAla was predicted to t into a binding cleft adjacent to the His-liganded catalytic Zn 2+ , thereby making favorable contacts, including potential H-bonds with the Gly236/Leu237 backbone of MMP-2 (Fig. 7B) and the Gly233/Leu234 backbone of MMP-9 (Fig. 7C). In contrast, in the complex with MMP-14, the bulkier Phe204 blocked access of the NCAA to the binding cleft occupied by HqAla in the complexes with MMP-2 and MMP-9 and resulted in local shifts in the backbones of both N-TIMP2-S69HqAla and MMP-14; in this case, there were no H-bonds of HqAla with MMP-14 (Fig. 7D). Overall, HqAla could thus form strong contacts with MMP-2 and MMP-9, but only minimal contact with MMP-14.
The modeling with the L-DOPA variants corroborated the impact of the MMP sequence differences at positions 196/193/204 (Fig. 8A). With MMP-2 and MMP-9, L-DOPA was predicted to t in close to the active site adjacent to the His-liganded catalytic Zn 2+ , unobstructed by Ala196 in MMP-2 or Pro193 in MMP-9. In the complex with MMP-2, L-DOPA69 was predicted to form a potential H-bond with the backbone of MMP-2 His233 (Fig. 8B), while with MMP-9, L-DOPA69 interactions included potential Hbonds with the backbone of Leu234 and side chain of His230 (Fig. 8C). In contrast, the Phe204 of MMP-14 prevented L-DOPA69 from accessing the binding cleft, and instead the modeling protocol predicted a rotamer pointing away from MMP-14 and potentially forming an intramolecular H-bond with Ser75 of N-TIMP2-S69DOPA (Fig. 8D). Overall, substitution of L-DOPA69 conferred novel strong interactions with either MMP-2 or MMP-9 but did not stabilize the interaction with MMP-14.

Discussion
This study presents a strategy for improving inhibitor speci city for individual enzymes within a homologous family, via mutagenesis of selected residues that participate in inhibitor-enzyme interactions. MMP family members share a common multi-domain structure, but exhibit differences in the subsites of their catalytic domains that lead to a variety of speci cities for different substrates and inhibitors. As MMPs are zinc-and calcium-dependent enzymes, a potential strategy for inhibiting MMPs could be to target these cations to disrupt their coordination by MMP residues. Since MMP subsites differ in their size and shape, the size and volume of amino acids within ligands and potential inhibitors, such as TIMP family members, will affect their a nity and selectivity towards different MMPs. In the current study, two approaches were combined to manipulate the a nity and selectivity of TIMP-2 for different MMPs. In the rst, the NCAAs L-DOPA and HqAla, which possess metal-binding capacity and have large side-chains, were incorporated into N-TIMP2 at selected positions with the potential to interact strongly with the Zn 2+ and Ca 2+ ions in the MMP catalytic domain. In the second approach, L-DOPA and HqAla were strategically placed to interact differently with distinctive subsites of different MMPs and hence to confer selectivity in binding and inhibition potencies. Incorporation of the bulky, polar, metal-binding NCAAs L-DOPA and HqAla at selected positions in N-TIMP2 did not improve its inhibition potency towards the examined MMPs-observations that highlight the important role played by the selected N-TIMP2 positions in the TIMP/MMP interactions. Nevertheless, the ndings that the inhibitory activity towards MMP-14 CAT was signi cantly impaired for all N-TIMP2-DOPA and N-TIMP2-HqAla variants, but was retained for MMP-2 ACT and MMP-9 CAT for most variants, emphasize the potential of these positions to alter N-TIMP2 selectivity towards different MMPs.
An examination of the different substitution positions within N-TIMP-2 provides explanations for our ndings. Position Ser2 of N-TIMP2, which is located in the N-terminal segment (residues Cys1-Pro5), is involved in the direct interaction between N-TIMP2 and the MMP catalytic pocket 13 and is therefore is intolerant of mutation to a bulkier residue, as we observed for the substitutions with either L-DOPA or HqAla and as was previously shown for other substitutions, such as S2E 38 and S2D 39 , at this position.
Our results are thus in line with these ndings, as Ser2 substitutions with both NCAAs led to signi cantly decreased a nities towards all tested MMPs, with a K i fold > 13, except for the retention of MMP-9 CAT inhibition by N-TIMP2-S2DOPA. Notably, N-TIMP2-S2HqAla showed > 2 orders of magnitude decrease in the a nity towards all three MMPs, suggesting that the bulky side chain of HqAla, compared to Ser, interferes with the MMP binding, probably due to steric hindrance.
Position Y36 of N-TIMP2 is located on the tip of the AB loop (residues D30-K41) and interacts with MMP-14 at a site that is distant from the catalytic pocket 40,41 . Different mutations (Y36F, Y36G and Y36W) led to decreased a nities towards MMP-14, exhibiting signi cant higher inhibition constants (K i -fold of 15 to 103) and lower association rate constants (K on -fold of 36 to 180), compared to N-TIMP2, whereas their a nities towards MMP-2 were maintained 42 . Our results extend these previous ndings, as N-TIMP2-Y36HqAla showed a 6.64-fold decreased a nity towards MMP-14 CAT , whereas it maintained its inhibitory potency towards both MMP-2 ACT and MMP-9 CAT , compared to N-TIMP2.
Positions S69 and A70 are located on the surface-exposed C-connector loop of N-TIMP2 (residues Ser68- Position L100 on N-TIMP2 is located on the EF loop between two beta-strands, sE and sF 37 . In this looppreviously identi ed as one of the N-TIMP2 binding sites for MMP-3 45 and MMP-14 -L100 is in close proximity to the MMP-14 catalytic Zn ion 44 , which may explain the observed reduction (by ~ 20-fold) in a nity towards MMP-14 CAT upon substitution of L100 with L-DOPA.
Our molecular modeling reveals how local sequence differences between the MMPs lead to differential susceptibility to inhibition by N-TIMP2 variants with insertion of the bulky HqAla or L-DOPA in position 69. Speci cally, MMP-14 is much less susceptible to inhibition by the variants as a consequence of deleterious steric interactions between HqAla or L-DOPA and MMP-14 Phe204, a position occupied by smaller residues in MMP-2 and MMP-9. Overall, this work suggests that different TIMP/MMP complexes have differential ability to tolerate the introduction of bulky residues within interface positions. In the absence of crystal structures, molecular dynamic simulations can be used to elucidate the molecular basis for these differences in selectivity.
In summary, in this study, the properties of HqAla and L-DOPA that shaped their differential interactions with the different MMPs can be attributed to their bulkiness and ability to form polar interactions, rather than to their known metal-binding capability. In the future, however, our approach might be extended to take advantage of metal coordination by metal-binding NCAAs at the interface of N-TIMP2 with its

Declarations
Availability of Data and Materials The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.